Abstract:At present, for business lightning positioning systems, the classification of lightning discharge types is mostly based on lightning pulse signal features, and there is still a lot of room for improvement. We propose a lightning discharge classification method based on convolutional encoding features. This method utilizes convolutional neural networks to extract encoding features, and uses random forests to classify the extracted encoding features, achieving high accuracy discrimination for various lightning discharge events. Compared with traditional multi-parameter-based methods, the new method proposed in this paper has the ability to identify multiple lightning discharge events and does not require precise detailed feature engineering to extract individual pulse parameters. The accuracy of this method for identifying lightning discharge types in intra-cloud flash (IC), cloud-to-ground flash (CG), and narrow bipolar events (NBEs) is 97%, which is higher than that of multi-parameter methods. Moreover, our method can complete the classification task of lightning signals at a faster speed. Under the same conditions, the new method only requires 28.2 μs to identify one pulse, while deep learning-based methods require 300 μs. This method has faster recognition speed and higher accuracy in identifying multiple discharge types, which can better meet the needs of real-time business positioning.

What problem does this paper attempt to address?

The problem this paper attempts to address is the inadequacy of existing commercial lightning location systems in classifying lightning discharge types. Currently, these systems primarily classify based on the characteristics of lightning pulse signals, but there is significant room for improvement in this approach. Specifically, the paper proposes a lightning discharge classification method based on convolutional encoded features, aiming to improve the classification accuracy of various lightning discharge events and also outperform traditional methods in terms of recognition speed. ### Background and Problem Description: 1. **Importance of Lightning Classification**: - Lightning is one of the most common discharge phenomena in the atmosphere, with approximately 30 to 100 lightning events occurring globally every second. - Lightning events are generally divided into two categories: Intra-cloud lightning (IC) and Cloud-to-ground lightning (CG). - Intra-cloud lightning accounts for about three-quarters of all lightning events, while cloud-to-ground lightning poses a serious threat to human survival and can lead to biological fatalities. 2. **Limitations of Existing Methods**: - The multi-parameter method is the most commonly used lightning classification method, characterizing electromagnetic field waveforms by extracting time-domain features of the waveform (such as amplitude ratio, fall time, rise time, and zero-crossing time). - Practical verification results show that the multi-parameter method has low accuracy in classifying lightning signals. For example, in the 2005 practical test of the National Lightning Detection Network (NLDN), about 54% of intra-cloud lightning was misclassified as cloud-to-ground lightning. - The classification accuracy of special intra-cloud lightning events (such as Narrow Bipolar Events, NBEs) is even lower, mainly affected by the multi-parameter classification method. 3. **Advantages of Deep Learning**: - Deep learning has made breakthrough progress in fields such as video recognition, audio analysis, and medical diagnosis. - Some studies have used Support Vector Machine (SVM) methods to classify very low-frequency lightning waveforms, achieving an accuracy rate of 97%. - The application of Convolutional Neural Networks (CNN) in lightning signal classification has also shown high accuracy, but the recognition efficiency is low, requiring high-performance hardware devices, making it difficult to meet real-time classification needs. ### Proposed Method in the Paper: - **Classification Method Based on Convolutional Encoded Features**: - Utilize Convolutional Neural Networks to extract encoded features and use Random Forest to classify the extracted encoded features. - This method can efficiently identify various lightning discharge events without the need for precise and detailed feature engineering to extract individual pulse parameters. - Achieves a recognition accuracy of 97% for Intra-cloud lightning (IC), Cloud-to-ground lightning (CG), and Narrow Bipolar Events (NBEs), higher than the multi-parameter method. - Faster recognition speed, taking only 28.2 microseconds to recognize a pulse, while deep learning-based methods require 300 microseconds. ### Summary: This paper proposes a lightning discharge classification method based on convolutional encoded features, aiming to improve the classification accuracy and recognition speed of lightning discharge types to better meet the needs of real-time business location.

A Lightning Classification Method Based on Convolutional Encoding Features

Classification of VLF/LF Lightning Signals Using Sensors and Deep Learning Methods

A Convolutional Neural Network for Classification of Lightning LF/VLF Waveform

Fast and Fine Location of Total Lightning from Low Frequency Signals Based on Deep-Learning Encoding Features.

Simultaneous Observations of Optical and Electrical Signals and Pattern Recognition in Lightning Flashes Classification

A Lightning Optical Automatic Detection Method Based on a Deep Neural Network

LD‐Net: A novel one‐stage knowledge distillation algorithm for lightning detection network

Monitoring Lightning Location Based on Deep Learning Combined with Multisource Spatial Data

Multi-scale Residual Transformer for VLF Lightning Transients Classification

A recognition method of electromagnetic pulse based on ensemble learning of adaptive signal decomposition features

An Efficient Compression Method for Lightning Electromagnetic Pulse Signal Based on Convolutional Neural Network and Autoencoder

LightNet+: A Dual-Source Lightning Forecasting Network with Bi-Direction Spatiotemporal Transformation

A spatio-temporal fusion deep learning network with application to lightning nowcasting

Lightning Performance Evaluation of Transmission Line Based on Data-Driven Lightning Identification, Tracking, and Analysis

Convective Storm VIL and Lightning Nowcasting Using Satellite and Weather Radar Measurements Based on Multi-Task Learning Models

Lightning Location and Peak Current Estimation From Lightning-Induced Voltages on Transmission Lines With a Machine Learning Approach

Detection Analysis and Fuzzy Recognition Investigation for Lightning Signal

Application of Ensemble Empirical Mode Decomposition in Low-Frequency Lightning Electric Field Signal Analysis and Lightning Location

A New Method of Three‐Dimensional Location for Low‐Frequency Electric Field Detection Array

A vision transformer for lightning intensity estimation using 3D weather radar

A physics-based ensemble machine-learning approach to identifying a relationship between lightning indices and binary lightning hazard